Biochemical characterization of nuclear pore complex protein gp210 oligomers

Abstract

The membrane-spanning glycoprotein gp210 is a major component of the nuclear pore complex. This nucleoporin contains a large cisternal N-terminal domain, a short C-terminal cytoplasmic tail, and a single transmembrane segment. We show here that dimers of native gp210 can be isolated from cell extracts by immunoprecipitation, and from purified rat liver nuclear envelopes by velocity sedimentation and gel filtration. Cross-linking of proteins in isolated membranes prior to solubilization dramatically increases the proportion of dimers. The dimers are SDS-resistant, as previously observed for some integral membrane proteins of cis-Golgi and plasma membrane proteins, including glycophorin A. Larger oligomers of gp210 can also be obtained by gel filtration and denaturing electrophoresis, but unlike the dimers are dissociated by reduction and heating in the presence of SDS. We propose that gp210 is organized into the pore membrane as a large array of gp210 dimers that may constitute a luminal submembranous protein skeleton.

Abbreviations

DMEM

Dulbecco modified Eagle's medium

NE

nuclear envelope.

DMA

dimethyl apimidate.

The nuclear membrane is constituted of three functionally and biochemically distinct domains, the inner, outer and pore membrane domains [1]. The outer nuclear membrane is associated with ribosomes with a protein composition identical to that of the rough ER. The inner nuclear membrane is associated with the nuclear lamina and chromatin with a distinct assortment of integral membrane proteins. Sharply bent nuclear pore membrane connects the inner and outer nuclear membranes and surrounds a multiprotein channel of 125 MDa, together forming the nuclear pore complex [1–3] that mediates the selective and bidirectional traffic of macromolecules into and out of the nucleus [4].

Two pore membrane proteins with single transmembrane domains, gp210 and POM121, have been identified and characterized at a molecular level in vertebrates [5,6]. Both gp210 and POM121 are type I integral membrane proteins with cytoplasmic C-terminal domains and N-terminal domains located in the lumen of the perinuclear space [7,8]. A remarkable difference between these two proteins is the size of their luminal and cytoplasmic (pore exposed) domains. Whereas the bulk of gp210 is located in the lumen of the NE, the majority of POM121, including a region containing FXFG repeats also found in other nucleoporins [9], is facing the cytoplasm.

The cytoplasmic C-terminal tail of gp210 contains 58 amino acids, while its luminal domain contains 1808 amino acids (including the signal sequence) with several N-linked oligosaccharides [5,7]. The large size and abundance of gp210, about 25 copies per nuclear pore complex representing 5–6 MDa [10], suggests that this protein is a major component of the luminal spoke domain of the nuclear pore complex [3,5].

The function(s) of gp210 in the nuclear envelope have not been elucidated. gp210 may have a structural role in anchoring the central framework of the pore complex to the membrane, although the identities of the putative ligands of gp210 in the nuclear pore complex remain unknown. Another potential function for gp210 may be the transmission of signals from the perinuclear cisterna to the nuclear pore complex by membrane-spanning segment and cytoplasmic domain [11]. gp210 has also been proposed to play a role in the membrane fusion event that takes place when a new nuclear pore complex is formed, because of the presence in the luminal domain of a 23-amino-acid hydrophobic peptide that might function as a potential fusion peptide [5].

It is believed that after integration into the ER membrane, the nuclear pore complex membrane proteins laterally diffuse through the lipid bilayer to their location within the pore membrane. Sorting signals for the targeting of gp210 and POM121 are contained in the pore exposed portion of the molecules [8,12], the transmembrane domain of gp210 being an additional sorting determinant [12]. These experiments suggested the specific homotypic or/and heterotypic lateral association of both proteins in the nuclear pore membrane. Here we provide evidence that gp210 is able to self-associate, forming homodimers that further associate into large arrays. We suggest that oligomerized gp210 dimers may organize a circular luminal skeleton for the pore nuclear membrane.

Experimental procedures

Reagents and antibodies

Chemical reagents were from Sigma unless otherwise indicated. Coomassie Brillant Blue and molecular mass markers were obtained from Biorad. Radioactive amino-acids were obtained from Amersham. Dystrophin (427 kDa) was obtained from the electric organ of Torpedo Marmorata[13,14]. Protein A–Sepharose and protein G–Sepharose were obtained from Sigma. Rabbit antibody C25 directed against a 25-amino-acid peptide from the C-terminus of gp210 has been described previously [15].

Isolation of rat liver nuclear envelopes

Rat liver nuclei were isolated from 300 g Sprague Dawley rats as described by Wozniak et al. [5], except that the buffer was 20 mm Tris/HCl (pH 7.5), and that iodoacetamide (25 mm) and dithiothreitol (1 mm) were present at the homogeneization and subsequent steps, respectively. Nuclear envelopes were prepared as described by Dwyer & Blobel [16] and modified by Wozniak et al. [5]. In some experiments, NE were directly solubilized in SDS sample buffer [17]. In other experiments, NE were extracted for 15 min with Triton X-100 at the indicated concentrations, in ice cold medium containing 25 mm Tris/HCl, pH 7.4, 0.3 m NaCl, 1 mm dithiothreitol, 0.5 mm 4-(2-aminoethyl)-benzenesulfonyl fluoride. After solubilization, samples were centrifuged for 15 min at 128 000 g in TL-100 centrifuge (Beckman), and the pellets were directly solubilized in SDS sample buffer. The supernatant fractions were mixed with the SDS sample buffer in order to keep the Triton X-100/SDS molar ratio identical (2%/0.05%) in each samples. The final volume of the pellet and supernatant samples were made identical before analysis by SDS/PAGE on 4% gels as described below. Alternatively, Triton X-100 supernatants were analyzed by velocity sedimentation or gel filtration as described below.

Immunoprecipitation

Metabolically labeled HeLa cells, grown on 35-mm dishes, were incubated for 1.5 h in methionine- and cysteine-free minimal essential medium and dialyzed serum, then in the same medium containing a mixture of ³⁵S-labeled methionine and cysteine (0.2–1 mCi·mL⁻¹) for 15–30 min. Cells were washed in NaCl/P_i, then extracted for 1 min at 4 °C on ice in 1 mL of a medium containing 20 mm Tris/HCl, pH 7.5, 3 mm MgCl₂, 1% Triton X-100, 1 mm dithiothreitol, and proteases inhibitors as above. After retrieval of this ER-enriched extract, dishes were washed in cold NaCl/P_i, then incubated for 10 min in 0.5 mL of 20 mm Tris/HCl, pH 7.5, 5 mm EDTA, 1% Triton X-100, 0.5 m NaCl, 1 mm dithiothreitol, and proteases inhibitors as above. Both lysates were sonicated, cleared by centrifugation at 18 000 g for 15 min, then processed for immunoprecipitation by affinity-purified C25 antibody [18]. All immunoadsorbant beads were washed as described [18] and antigens were eluted by heating the beads for 15 min at 70 °C in SDS sample buffer containing 60 mm dithiothreitol. In some experiments, dithiothreitol was omitted as indicated.

Gel electrophoresis and immunoblotting

Protein electrophoresis was performed either on 5–20% linear gradient polyacrylamide gels, or on 4, 6, 8 or 10% continuous gels according to Laemmli [17]. Samples were incubated at 70 °C for 15 min in the presence of 100 mm dithiothreitol, unless otherwise indicated. Wet electrophoretic transfer of rat liver NE proteins and immunoblotting analysis were performed according to standard methods [19]. After a final incubation with either horse-radish peroxidase conjugated anti-(rabbit IgG) Ig (Promega), blots were revealed by ECL (Amersham) according to the manufacturer' instructions.

Proteolysis of gp210

[³⁵S]Methionine/cysteine-labeled HeLa cell extracts were immunoprecitated by C25 antibodies as described above. Immunoprecipitate was resolved by SDS/PAGE on a 6% gel, and the region of the gel corresponding to the 200- to 400-kDa proteins was excised and submitted to mild proteolysis digestion with Staphylococcus aureus V8 protease (Boehringer) according to Cleveland et al. [20], using 2 µg·mL⁻¹ of enzyme and a 120-min incubation at 18 °C. Separation of the proteolyzed fragments was achieved on a 15% SDS polyacrylamide gel. Revelation of the labelled peptides was performed by fluorography [21].

Sucrose gradient sedimentation and gel filtration

For sedimentation experiments, NE were solubilized as described above in the presence of 2% Triton X-100 and 1 m NaCl and the soluble fraction was loaded on the top of continuous 5–20% sucrose (w/w) linear gradients in the same medium. Gradients were centrifuged in a SW41 rotor at 200 000 g for 14–16 h. Fractions of 0.8 mL were collected from the bottom, precipitated by 10% trichloroacetic acid, before solubilization in SDS sample buffer. For gel filtration experiments, NE were incubated for 1 h on ice in a medium containing 25 mm Tris/HCl, pH 7.4, 2% Triton X-100, 0.3 m NaCl, 1 mm dithiothreitol, 0.5 mm 4-(2-aminoethyl)-benzenesulfonyl fluoride, then centrifuged for 15 min at 100 000 g in a Beckman TL100 centrifuge and the soluble fraction was loaded onto an HPLC-connected Superose 6 column (Pharmacia) previously equilibrated in 25 mm Tris/HCl, pH 7.4, 0.1% Triton X-100, 150 mm NaCl, 0.5 mm 4-(2-aminoethyl)-benzenesulfonyl fluoride, 1 mm dithiothreitol. The sample was run at 0.5 mL·min⁻¹ flow rate and 1-mL fractions were collected. Each fraction was solubilized in 2 × SDS sample buffer, resolved on a 4% SDS/PAGE and finally analyzed by immunoblotting as described below. Blue Dextran (2 MDa), thyroglobulin (≈ 680 kDa) and standard molecular mass markers (Combithek, Boehringer Mannheim) were used as indicators of the elution size.

Cross-linking experiments

Dimethyl apidimidate was made up freshly in 100 mm triethanolamine, pH 8.5, then used at 200 mm. NEs were incubated for 30 min at room temperature with dimethyl apidimidate; the cross-linker was then neutralized by 200 mm glycine, before solubilization of the NEs in 2 × SDS sample buffer. Proteins in the samples were resolved by SDS/PAGE on a 4% gel, then analyzed by immunoblotting using C25 antibodies.

Results

Gp210 forms dimers that are stable in SDS

As previous data have suggested that gp210 can generate SDS-resistant dimers [22], we studied if dimers of endogenous gp210 could be obtained from lysates of HeLa cells and from isolated rat liver NEs. HeLa cell proteins were metabolically labeled by incorporation of [³⁵S]methionine and cysteine for 30 min, then lysed sequentialy in the presence of dithiothreitol with 1% Triton X-100, first in the absence of salt to obtain an ER-enriched fraction, then in the presence of 0.5 m NaCl to extract pore-associated gp210 [10]. Both fractions were immunoprecipitated with C25 antibodies directed against gp210, then analyzed by SDS/PAGE on a 6% gel. Data of Fig. 1 (lanes 2 and 3) show that gp210 was immunoprecipitated together with a component with a molecular mass close to 400 kDa, as judged from its comigration with 427-kDa dystrophin (data not shown). As the ≈ 400 kDa mass of the larger immunoreactive component suggested the formation of an homodimer of gp210, this hypothesis was explored by peptide mapping. Data show that high molecular mass peptides generated by the protease digestion of the ≈ 400-kDa component (Fig. 1B, lane 4) were similar to that generated by gp210 (Fig. 1B, lane 5, arrowheads), while the low molecular mass peptide products were slightly different. In the peptide map generated by the ≈ 400-kDa component, there was no superposition of a new peptide map to the gp210 peptide map suggesting that the ≈ 400-kDa component is a dimer of gp210, the differences in the size of the low molecular mass peptide products between the two maps being interpreted as a different accessibility of the protease to the cleavage sites in the monomer and dimer. As a unique component of ≈ 7 kb has been identified by Northern blotting from polyA⁺ -selected RNA using a gp210 DNA probe [5], the existence of a ≈ 400-kDa translation product for gp210 can be ruled out. Altogether, these data support the fact that the ≈ 400-kDa component that reacts with the antibodies directed against gp210 is a SDS-resistant dimer of gp210. Moreover, these results suggest that the dimers are formed in the ER before the assembly of gp210 into the nuclear pore complex. The dimer/monomer ratio was higher in the nuclear pore complex than in the ER fraction, perhaps due to a stabilization of the dimers in the nuclear pore complex by interaction with other dimers and/or with components of the central core of the nuclear pore complex.

Figure 1

SDS-resistant dimers of gp210 can be isolated from cell extracts. HeLa cells were labeled by a 30-min incorporation of [³⁵S]methionine/cysteine, and then lysed sequentially with 1% Triton X-100, first in the absence of salt, then in the presence of 0.5 m NaCl. Proteins from both lysates were immunoprecipitated with anti-gp210 Ig, and then resolved by SDS/PAGE on a 6% gel followed by fluorography (A). Lane 1 corresponds to an immunoprecipitation of the total extract by preimmune serum; lanes 2 and 3 show proteins immunoprecipitated from low salt extract and high salt extract, respectively. Numbers on the left correspond to molecular masses in kDa. Note that a ≈ 400-kDa component is immunoprecipitated together with gp210 (200 kDa). gp210 and ≈ 400-kDa component obtained by immunoprecipitation and resolved by denaturing electrophoresis as in (A), were cleaved in gel by V8 protease [20]; digested peptides were then separated on a 15% SDS-polyacrylamide gel and revealed by fluorography (B). Proteolytic peptides of gp210 are shown in lane 5 and that of the cross-reactive ≈ 400-kDa component in lane 4. Note the similarity in size of the high molecular mass peptides obtained from both proteins (arrowheads).

The properties of gp210 dimers were further explored after extraction from purified NEs. NEs isolated from rat livers in the presence of iodoacetamide at the tissue homogenization step and of dithiothreitol at all subsequent steps, were solubilized in SDS electrophoresis sample buffer under different conditions, then resolved by electrophoresis on a 4% linear polyacrylamide gel, and finally analyzed by immunoblotting (Fig. 2A). Without heating of the samples, several immunoreactive components of 200 kDa, ≈ 400 kDa, and even higher molecular masses were detected, either in the absence or presence of 100 mm dithiothreitol (Fig. 2A, lanes 1 and 2). Heating of the sample for 15 min at 70 °C resulted in a decrease in intensity of gp210 monomeric and dimeric signals and in the disappearance of higher molecular mass signals (Fig. 2A, lane 3). Heating of the sample at 95 °C for 5 min abolished the signals for dimeric gp210, and severely weakened the intensity of the monomeric signal (Fig. 2A, lane 4), probably through an irreversible aggregation of gp210, as previously reported also for other transmembrane proteins [23]. No modification of the monomer to dimer ratio, as quantified by scanning the intensity of the ECL signals, was observed after reduction of the concentration of SDS (from 2 to 0.02%), or after a 10-fold dilution of protein concentration (data not shown).

Figure 2

SDS-resistant dimer of gp210 can be obtained from isolated NE. Rat liver NE proteins were solubilized under different experimental conditions, then resolved by 4% SDS/PAGE, and finally analyzed by immunoblotting using anti-gp210 Ig. NEs were solubilized (A) either in 2 × SDS sample buffer without dithiothreitol and heating (lane 1), or in 2 × SDS sample buffer in the presence of 100 mm dithiothreitol (lane 2), or as in lane 2 before heating at 70 °C for 15 min (lane 3), or treated as in lane 2 but heated at 95 °C for 15 min (lane 4). Note the presence in lane 1 and 2 of oligomers of a size higher than dimers (400 kDa). Rat liver NEs were extracted (B) in the presence of 0.3 m NaCl with either 0.5% Triton X-100 (lanes 1 and 2), or 0.9% Triton X-100 (lanes 3 and 4). Soluble (lanes 1 and 3) and insoluble (lanes 2 and 4) material were isolated by sedimentation, then resolved by immunoblotting as in lane 3 of (A). Rat liver NEs were incubated for 2 h in the presence of 200 mm of the cross-linking reagent dimethyl apidimidate, then quenched by glycine, were finally solubilized in SDS (C). Proteins resolved by electrophoresis on a 4% gel, were analyzed by immunoblotting as in lane 3 of (A). Note that, compared to the control (lane 1), the signal for the 400 kDa component is strongly increased in dimethyl apidimidate-treated NEs (lane 2). Different mobilities of the 400 kDa component in (A), (B), and (C), are due to the fact that this panels correspond to independant experiments. S, supernatant; P, pellet; DTT, dithiothreitol.

Differential extraction of monomeric and dimeric forms of gp210

NEs were solubilized at different concentrations of Triton X-100 in the presence of 0.3 m NaCl, and the soluble and insoluble fractions were isolated by sedimentation before analysis by immunoblotting. At a concentration of 0.5% Triton X-100, most of gp210 was solubilized, however, almost exclusively under its monomeric form (Fig. 2B, lane 1). The dimeric form of gp210 was found in the pellet, together with a minor fraction of the monomeric form (Fig. 2B, lane 2). At a concentration of 0.9% Triton X-100, all the monomeric fraction of gp210 was solubilized, together with a large fraction of the SDS-resistant dimeric form of gp210 (Fig. 2B, lane 3). However, even under these stringent conditions, dimers of gp210 were still found in the insoluble fraction (Fig. 2B, lane 4). The differential extraction of both forms of gp210 by increasing concentrations of Triton X-100 at a constant salt concentration, suggests that either the dimer is more strongly bound than the monomer to another pore complex component or that it is located in a specialized membrane compartment that is very resistant to solubilization by Triton X-100.

Cross-linking experiments

To further rule out the possibility that gp210 dimers may be produced by nonspecific aggregation after membrane solubilization, NE were treated with the bifunctional cross-linker reagent dimethyl apidimidate before solubilization, then analyzed by denaturing electrophoresis. The data in Fig. 2C show that a 2-h incubation of NE with dimethyl apidimidate resulted in a decrease of the signal for monomeric gp210 with a simultaneous increase in the signal referring to the dimeric and higher forms of gp210. Therefore, we suggest that gp210 dimers preexist in the isolated membranes and are not generated after membrane solubilization.

Isolation of gp210 monomers and dimers by velocity sedimentation and gel filtation

Proteins solubilized from rat liver NEs in 0.5% Triton X-100 and 0.5 m NaCl were analyzed by velocity sedimentation on a continuous sucrose gradient containing 1% Triton X-100 and 1 m NaCl. Data of Fig. 3A show that the monomeric form of gp210 was sedimenting at ≈7S, which corresponds to the protein marker aldolase (7.4S). The dimer of gp210 was also present under these experimental conditions, sedimenting with a slightly higher velocity.

Figure 3

Separation of monomeric and dimeric forms of gp210 by sucrose velocity gradient sedimentation and gel chromatography. Rat liver NEs were solubilized in 2% Triton X-100 and 1 m NaCl, then sedimented at 200 000 g for 16 h on a 5–20% continuous sucrose gradient (A). Proteins in the fractions were resolved on a 5–20% continuous SDS-polyacrylamide gel, then analyzed by immunoblotting using anti-gp210 Ig. Note that the peak of monomeric gp210 (200) sediments at ≈ 7S (fraction 11), while the dimeric form (400) sediments faster (peak in fraction 10). Rat liver NE proteins solubilized in 2% Triton X-100 and 0.3 m NaCl, were analyzed by HPLC on a Superose 6 column, then fractions (Fr.) were resolved by SDS/PAGE on a 5–20% linear polyacrylamide gel (B) and revealed by silver staining (upper panel). Note that a ≈ 200-kDa protein sedimenting in fractions 9–13 is separated from other NE proteins present in further fractions. This 200-kDa component was identified as gp210 by immunoblotting (lower panel). Numbers on the left refer to molecular mass markers used during electrophoresis. Numbers below the figure are molecular mass values deduced from the calibration of the column. Column fractions obtained from an experiment similar to that of (B) were resolved by electrophoresis on a 4% gel, and then analyzed by immunoblotting using anti-gp210 Ig (C). Note that gp210 peak is composed of two peaks enriched in gp210 dimers (400) and monomers (200), respectively.

For chromatographic analysis, NE proteins solubilized in 2% Triton X-100 and 0.3 m NaCl were resolved on a Superose 6 column equilibrated in 0.1% Triton X-100, 0.15 m NaCl, and 1 mm dithiothreitol. Proteins present in the 1-mL fractions were analyzed by denaturing electrophoresis on a polyacrylamide gel gradient, followed by silver staining (Fig. 3B, upper panel). Figure 3B reveals a 200-kDa protein as the major component present in fractions corresponding to macromolecules eluted within a range of molecular masses from 2 to 0.8 MDa (fractions 9–13), all other solubilized proteins being eluted below 500 kDa (15–19). This 200-kDa component was identified as gp210 by immunoblotting (Fig. 3B, lower panel). When the high molecular mass peak was analyzed on a 4% gel under denaturing conditions (Fig. 3C), it was shown to be composed of two gp210 fractions. One oligomer fraction peaking at ≈ 1.55 MDa was composed of SDS-resistant gp210 dimers, another fraction peaking at ≈ 1.1 MDa was composed of gp210 monomers.

These data show that, under the stringent conditions generated by high concentrations of Triton X-100 and salt and the strong shearing force of velocity sedimentation, gp210 is both monomeric and dimeric. Under the less stringent conditions generated by a lower concentration of detergent and salt, and the low shearing force of gel filtration chromatography, polydisperse oligomers of gp210, some of them reaching the size of an octamer, were formed. These oligomers were dissociated by SDS into monomers and dimers.

Discussion

Nucleoporin gp210 solubilized by nonionic detergents from ER- and NE-enriched fractions of cells in culture, was immunoprecipitated as a mix of monomers and SDS-resistant dimers. The same SDS-resistant dimers were observed by immunoblotting after direct solubilization of rat liver NEs in SDS. Several lines of evidence support the idea that SDS-resistant gp210 dimers are preexisting in NEs before addition of the detergent and are not formed secondarily in the electrophoresis buffer: (a) neither dilution of the protein nor changes in SDS concentration modified the dimer/monomer ratio; (b) this ratio was strongly enhanced when NE solubilization was performed after treatment of the isolated NEs by cross-linking agents; (c) dimeric and monomeric gp210 populations were differentially extracted from NEs by a nonionic detergent, the dimeric fraction requiring more stringent conditions; and (d) both forms of gp210 can be isolated by velocity sedimentation and gel filtration after extraction from NEs by nonionic detergent.

Dimers of other transmembrane proteins have also been shown to exhibit SDS-resistance. SDS-resistance of dimers of zeta-chain of T-cell receptor [24] and CD8α from the plasma membane of cytotoxic T cells [25] is due to disulfide bridges. As reductive reagents were present at all steps of cellular fractionation and in electrophoresis sample buffer, it is unlikely that this mechanism can explain the SDS-resistance of gp210 dimers. By contrast with CD8α and zeta-chain dimers, SDS-resistance of glycophorin A dimers from red cell plasma membrane is due to features of its transmembrane domain [26–29]. A different mechanism has been proposed for SDS-resistant dimerization of M glycoprotein of avian coronavirus (IBV) in the cis-Golgi [30,31]. In this system, while the specificity of dimerization is dictated by the sequence of the membrane-spanning region, the cytoplasmic tail of the protein is necessary for SDS-resistance [32].

Oligomers of gp210, reaching the size of tetramers and octamers were observed following solubilization of NEs in SDS and denaturing electrophoresis. However, unlike the dimers, these oligomers did not resist heating of the sample in the presence of dithiothreitol. Large, SDS-sensitive oligomers of similar size were also obtained by gel filtration chromatography of Triton-solubilized nuclear envelope proteins. The self-association of gp210 under these experimental conditions was remarkable as it allowed a one-step isolation of gp210 from other peripheral and integral NE proteins.

Owing to structural features similar to those of viral fusion proteins, gp210 has been suggested to play a role in the membrane fusion event that takes place during pore biogenesis [5,7]. These features include glycosylation, the presence of a potential fusion peptide in the large luminal domain of the protein, the presence of a unique transmembrane domain, and the ability to form dimers and multimers [33–35]. For influenza virus hemagglutinin, a critical surface density and vicinity is required for the induction of membrane fusion [36]. In this context it is of interest that a pool of easily extractable gp210 was found in the NE, which may be free to diffuse laterally in the ER and participate in the triggering of the fusion of the outer and the inner nuclear membranes.

Altogether the data obtained in this study combined with the data of Wozniak & Blobel on the sorting of gp210 [12], suggest that a fraction of gp210 molecules is present in the pore membrane as dimers, probably maintained by homotypic interactions involving the transmembrane domain and possibly the C-terminal domain [12]. Organized by interaction with the pore complex core octamer, these dimers could form a large circular multimer that would surround the pore complex on the luminal side of the pore membrane. The minimal number of gp210 dimers required in this model would be of 8, a number compatible with the 16–25 molecules estimated by Gerace et al. [10] to be present per nuclear pore complex.

Acknowledgements

We gratefully acknowledge Drs R. W. Wozniak, G. Migliaccio, P. Engel and P. Cosson for the gift of reagents. We thank Drs P. Nicolas, J-L. Popot, V. Doye, and M. Lohka for helpful discussions. We thank Ms M. Barre for the artwork. This work was supported by grant 9227 from Association pour la Recherche contre le Cancer (ARC). P. M. was supported by fellowships of the Sigrid Juselius and Finnish Cultural foundations, R. B. was supported by la Fondation pour la Recherche Médicale (FRM) et la Société de Secours des Amis de la Science (SSAS).

Footnotes

*Present address: IDIBAPS, Hospital Clinic, Universitat de Barcelona, Villaroel 170, 08036, Barcelona, Spain.

†Present address: Department of Medical Chemistry, Institute of Biomedicine, POB 8 (Siltavuorenpenger 10 a), FIN-00014 University of Helsinki, Finland.